Measuring volunteered geodata performance in the text-to-GPS linkage of field-captured locality records

GIS applications that link natural language text to geographic space using gazetteers are essential for managing spatial data in archived geological records. However, these applications face limitations due to limited gazetteer scope of coverage. Crowdsourced gazetteers often have global coverage, making them excellent reference data for resolving spatial details in records, including converting textual descriptions of locations to GPS features. This can be useful for rectifying missing GPS information in field-captured geological records, especially those obtained from remote and hard-to-reach areas. However, accurately transforming location descriptions in text to GPS coordinates is challenging, and reference data quality can be crucial in minimizing errors and uncertainties. A list of mineral specimen localities referencing geological sampling sites in the Northwest Territories and Nunavut were geoparsed using Geonames and OpenStreetMap geocoders and match rates, positional accuracy, and lexical similarity were quantified to assess performance.

使用地名词典（gazetteer）将自然语言文本与地理空间关联起来的地理信息系统（GIS）应用，对于归档地质记录的空间数据管理至关重要。然而此类应用受限于地名词典的覆盖范围，存在一定局限性。
众包地名词典通常具备全球覆盖范围，可作为解决记录中空间细节问题的优质参考数据，包括将位置的文本描述转换为全球定位系统（GPS）特征。该功能可用于补全野外采集的地质记录中缺失的GPS信息，尤其适用于偏远难达区域获取的地质记录。
然而，将文本中的位置描述精准转换为GPS坐标仍颇具挑战，参考数据的质量对于最小化误差与不确定性至关重要。
本研究针对一份指向加拿大西北地区与努纳武特地区地质采样点的矿物标本点位列表，采用地名数据库（Geonames）与开放街道地图（OpenStreetMap）地理编码器开展地理解析，并通过量化匹配率、位置精度与词汇相似度来评估其性能。